quote: Original post by TerranFury
In game AI the main point is to be believable.

In your opinion. This wouldn''t necessarily be the case in many games. A potentially better goal would be to make a challenging opponent.

quote: This brings me to my point: What we want in game AI is a response which is not deterministic given data available to the player .

Again, this isn''t necessarily true. One reason is that making the responses partly random will allow players to save and reload in order to beat the opponent.

[ MSVC Fixes | STL | SDL | Game AI | Sockets | C++ Faq Lite | Boost | Asking Questions ]

Totally agree with TerranFury - it doesn''t really matter with regard to AI.

Newt - the brain might actually be a little bit random (I think Penrose tried to prove this but it''s doubtful that he was successful), and puzzler has pointed out that you can actually get pretty good random generators in any case. There is such a thing as random, which is why we have a word for it. I''m not sure that you can truly ''replay'' any quantum-level event in any case.

On the random numbers issue, i guess it depends on what definition of random you are using. Usually random means some degree of unpredictability. For example, lets say you have a dice for which there is an exactly equal chance of 1/6 for any of the results and that it is completely impossible to predict the result. Some might say that this is random but it isnt exactly, lets say you have a set of results from dice rolls - 14253, now the last number statistically has a high chance of being six, in a truly random set the chance for the next number in the set being six would have to be the same chance of it being five or three. Instead it becomes more predictable and not infact random. I read a paper on this issue before and barring quantum interactions that Im not entirely knowledgeable on, random numbers are not infact random because they rely on certain rules to be generated. With the dice example the rules are that the numbers have to be within one and six, and they have to have statistically about the same number of each. These rules make ''random'' numbers predictable and therefor not random at all.

Predictable? So you know what they are going to role next on a die? Cool...

BTW Rolling a die isn''t a quantum function... Radioactivity is...

I never said rolling a dice was, I said ''barring Quantum Functions which im not too knowledgeable on''.

quote: Original post by KingMolson
Some might say that this is random but it isnt exactly, lets say you have a set of results from dice rolls - 14253, now the last number statistically has a high chance of being six, in a truly random set the chance for the next number in the set being six would have to be the same chance of it being five or three. Instead it becomes more predictable and not infact random.

this is simply not true. if it is truly random (that theoretical "fair die") then there is exactly 1/6 chance that the next roll will be a six. the dice don''t care what you expect based on statistics.
if you flip a coin 1000 times and get heads 1000 times, there is still a 50% chance you will get heads next time (unless the coin is in fact not "fair"). this is the first thing they teach you in probability.

Yep, Krez is right.

I don''t understand what this discussion about randomness is all about. Who cares if the choices of the AI will be different? MY choices remain the same, (like in deciding to do tank drop everytime in starcraft, or planning my day) at least, until I learn something new. If there is an incentive to having a different decision each time, then it should be part of the decision process that weighes the options, and then that would remain the same. Why would the randomness be so important in AI then?

By the way, I''ve been thinking about True AI, and if you can make it, why would you want to make an amusing robot? True AI will be able to do any given enough time. Time, to computers, is inversly proportional to the computing power, using all the computer chips together would be able to bring world peace. Maybe we need better computer opponents in our games though.

Where is your artificial imagination message? I thought imagination only consisted of building theoretical situations based on desires and creativity.

Thats a good point Krez, I hadnt thought of that. If you think about it, if it didnt statistically throwout about an equal number of each then how would it be fair?

With regards to Randomness, I think of the definition thus:

Think of time as a tape being played in your vcr. Say you watch a ''random'' event take place. If you then rewound the tape and played the event again, and if the event was truly random, there would be a *chance* of a different outcome (you''d have to keep rewinding back and watching until something different *did* happen in order to prove it though, so if you''re wrong you could be there for ever :p ).

ie, given a certain situation, there are two or more *truly* possible outcomes. (not simply possible in the sense that you don''t know which one will come to pass)

- Note that true randomness is practically impossible to prove in the real world, unless someone figures out a way to rewind time (or somehow tells all the atoms in the universe to go back to where they were before, so you can do the experiment again).

"What the imagination seizes as beauty must be truth - whether it existed before or not." - John Keats